Scholars@Duke
Alan E. Gelfand

Alan E. Gelfand

James B. Duke Distinguished Professor Emeritus of Statistical Science
Statistical Science
alan@stat.duke.edu
Box 90251, Durham, NC 27708-0251
223A Old Chem Bldg, Durham, NC 27708

Scholarly Works - Conferences

Maximum weight matching using odd-sized cycles: Max-product belief propagation and half-integrality

Conference IEEE Transactions on Information Theory · March 1, 2018 We study the maximum weight matching (MWM) problem for general graphs through the max-product belief propagation (BP) and related Linear Programming (LP). The BP approach provides distributed heuristics for finding the maximum a posteriori (MAP) assignment ... Full text Cite

Hierarchical spatio-temporal modeling of resting state fMRI data

Conference Springer Proceedings in Mathematics and Statistics · January 1, 2018 In recent years, state of the art brain imaging techniques like Functional Magnetic Resonance Imaging (fMRI), have raised new challenges to the statistical community, which is asked to provide new frameworks for modeling and data analysis. Here, motivated ... Full text Cite

Geographic segmentation via latent poisson factor model

Conference Wsdm 2016 Proceedings of the 9th ACM International Conference on Web Search and Data Mining · February 8, 2016 Discovering latent structures in spatial data is of critical importance to understanding the user behavior of locationbased services. In this paper, we study the problem of geographic segmentation of spatial data, which involves dividing a collection of ob ... Full text Cite

Markov-modulated marked poisson processes for check-in data

Conference 33rd International Conference on Machine Learning Icml 2016 · January 1, 2016 We develop continuous-time probabilistic models to study trajectory data consisting of times and locations of user 'check-ins'. We model the data as realizations of a marked point process, with intensity and mark-distribution modulated by a latent Markov j ... Cite

A graphical transformation for belief propagation: Maximum Weight Matchings and odd-sized cycles

Conference Advances in Neural Information Processing Systems · January 1, 2013 Max-product 'belief propagation' (BP) is a popular distributed heuristic for finding the Maximum A Posteriori (MAP) assignment in a joint probability distribution represented by a Graphical Model (GM). It was recently shown that BP converges to the correct ... Cite

Loop calculus and bootstrap-belief propagation for perfect matchings on arbitrary graphs

Conference Journal of Physics Conference Series · January 1, 2013 This manuscript discusses computation of the Partition Function (PF) and the Minimum Weight Perfect Matching (MWPM) on arbitrary, non-bipartite graphs. We present two novel problem formulations-one for computing the PF of a Perfect Matching (PM) and one fo ... Full text Cite

Belief propagation for linear programming

Conference IEEE International Symposium on Information Theory Proceedings · January 1, 2013 Belief Propagation (BP) is a popular, distributed heuristic for performing MAP computations in Graphical Models. BP can be interpreted, from a variational perspective, as minimizing the Bethe Free Energy (BFE). BP can also be used to solve a special class ... Full text Cite

Generalized belief propagation on tree robust structured region graphs

Conference Uncertainty in Artificial Intelligence Proceedings of the 28th Conference Uai 2012 · December 1, 2012 This paper provides some new guidance in the construction of region graphs for Generalized Belief Propagation (GBP). We connect the problem of choosing the outer regions of a Loop- Structured Region Graph (SRG) to that of finding a fundamental cycle basis ... Cite

A cluster-cumulant expansion at the fixed points of belief propagation

Conference Uncertainty in Artificial Intelligence Proceedings of the 28th Conference Uai 2012 · December 1, 2012 We introduce a new cluster-cumulant expansion (CCE) based on the fixed points of iterative belief propagation (IBP). This expansion is similar in spirit to the loop-series (LS) recently introduced in [1]. However, in contrast to the latter, the CCE enjoys ... Cite

Integrating local classifiers through nonlinear dynamics on label graphs with an application to image segmentation

Conference Proceedings of the IEEE International Conference on Computer Vision · December 1, 2011 We present a new method to combine possibly inconsistent locally (piecewise) trained conditional models p(y αx α) into pseudo-samples from a global model. Our method does not require training of a CRF, but instead generates samples by ... Full text Cite

Pushing the power of stochastic greedy ordering schemes for inference in graphical models

Conference Proceedings of the National Conference on Artificial Intelligence · November 2, 2011 We study iterative randomized greedy algorithms for generating (elimination) orderings with small induced width and state space size - two parameters known to bound the complexity of inference in graphical models. We propose and implement the Iterative Gre ... Cite

Stopping rules for randomized greedy triangulation schemes

Conference Proceedings of the National Conference on Artificial Intelligence · January 1, 2011 Many algorithms for performing inference in graphical models have complexity that is exponential in the treewidth - a parameter of the underlying graph structure. Computing the (minimal) treewidth is NP-complete, so stochastic algorithms are sometimes used ... Full text Cite

On herding and the perceptron cycling theorem

Conference Advances in Neural Information Processing Systems 23: 24th Annual Conference on Neural Information Processing Systems 2010, NIPS 2010 · December 1, 2010 The paper develops a connection between traditional perceptron algorithms and recently introduced herding algorithms. It is shown that both algorithms can be viewed as an application of the perceptron cycling theorem. This connection strengthens some herdi ... Cite

Beem : BBucket Elimination with external memory

Conference Proceedings of the 26th Conference on Uncertainty in Artificial Intelligence Uai 2010 · January 1, 2010 A major limitation of exact inference algorithms for probabilistic graphical models is their extensive memory usage, which often puts real-world problems out of their reach. In this paper we show how we can extend inference algorithms, particularly Bucket ... Cite

Performance evaluation of decentralized estimation systems with uncertain communication

Conference 2009 12th International Conference on Information Fusion Fusion 2009 · January 1, 2009 An approach for evaluating the performance of decentralized estimation systems under non-ideal communi-cations is presented. Recent studies have shown that trans-mission disruptions occur frequently in tactical wireless net-works due to a variety of unpred ... Cite

Advanced algorithms for distributed fusion

Conference Proceedings of SPIE the International Society for Optical Engineering · June 5, 2008 The US Military has been undergoing a radical transition from a traditional "platform-centric" force to one capable of performing in a "Network-Centric" environment. This transformation will place all of the data needed to efficiently meet tactical and str ... Full text Cite

Statistical comparison of a hybrid approach with approximate and exact inference models for fusion 2+

Conference Proceedings of SPIE the International Society for Optical Engineering · November 15, 2007 One of the greatest challenges in modern combat is maintaining a high level of timely Situational Awareness (SA). In many situations, computational complexity and accuracy considerations make the development and deployment of real-time, high-level inferenc ... Full text Cite

Suppression and failures in sensor networks: A Bayesian approach

Conference 33rd International Conference on Very Large Data Bases VLDB 2007 Conference Proceedings · January 1, 2007 Sensor networks allow continuous data collection on unprecedented scales. The primary limiting factor of such networks is energy, of which communication is the dominant consumer. The default strategy of nodes continually reporting their data to the root re ... Cite

Introduction of the hybrid inference tool (HIT)

Conference Fusion 2007 2007 10th International Conference on Information Fusion · January 1, 2007 The construction of belief networks is a widely used methodology for high level fusion modeling. While some of the components of a belief network deal with ambiguous (probabilistic) data, others may deal with vague (possibilistic) data. Given the need to r ... Full text Cite

From data reverence to data relevance: Model-mediated wireless sensing of the physical environment

Conference Lecture Notes in Computer Science Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics · January 1, 2007 Wireless sensor networks can be viewed as the integration of three subsystems: a low-impact in situ data acquisition and collection system, a system for inference of process models from observed data and a priori information, and a system that controls the ... Full text Cite

Data-driven processing in sensor networks

Conference Cidr 2007 3rd Biennial Conference on Innovative Data Systems Research · January 1, 2007 Wireless sensor networks are poised to enable continuous data collection on unprecedented scales, in terms of area location and size, and frequency. This is a great boon to fields such as ecological modeling. We are collaborating with researchers to build ... Cite